DeepMatrixanim

Animations for understanding deep learning through matrix and vector operations.

Prerequisites

• Windows 11
• uv
• MiKTeX with latex, pdflatex, and dvisvgm available on PATH
• FFmpeg with ffmpeg available on PATH

Setup

$env:UV_CACHE_DIR = Join-Path (Get-Location) ".uv-cache"
uv sync --group dev

Verify

$env:UV_CACHE_DIR = Join-Path (Get-Location) ".uv-cache"
uv run pytest -v
uv run manim --version

Render The Demos

$env:UV_CACHE_DIR = Join-Path (Get-Location) ".uv-cache"
uv run manim --custom_folders -ql scenes/mlp_forward_demo.py MLPForwardDemo
uv run manim --custom_folders -ql scenes/linear_layer_space_demo.py LinearLayerAsSpaceTransform
uv run manim --custom_folders -ql scenes/relu_space_folding_demo.py ReLUAsSpaceFolding
uv run manim --custom_folders -ql scenes/representation_space_demo.py RepresentationSpaceDecisionBoundary
uv run manim --custom_folders -ql scenes/iris_mlp_training_demo.py IrisMLPBeforeAfterTraining
uv run manim --custom_folders -ql scenes/rotary_position_embedding_space_demo.py RotaryPositionEmbeddingSpaceDemo
uv run manim --custom_folders -ql scenes/rope_relative_dot_product_demo.py RoPERelativeDotProductDemo
uv run manim --custom_folders -ql scenes/rope_attention_matrix_demo.py RoPEAttentionMatrixDemo

The project manim.cfg keeps rendered videos for all scene files in one quality folder when --custom_folders is used, such as media/videos/480p15/.